Visual Speech Classification on VSC
58nAUC (0-5 pp budget)GACL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GACLModel=Qwen2.5-Omni2026.06 | 58 | 71.6 | |
| AADModel=Qwen2.5-Omni2026.06 | 48.4 | 56.6 | |
| GACLModel=Qwen2-Audio-Inst.2026.06 | 39.1 | 51.2 | |
| GACLModel=Qwen3-Omni2026.06 | 36.2 | 42.5 | |
| ACDModel=Qwen2.5-Omni2026.06 | 29.4 | 37 | |
| GACLModel=Voxtral-Small2026.06 | 25 | 28.1 | |
| AADModel=Qwen2-Audio-Inst.2026.06 | 18.4 | 29.4 | |
| GACLModel=Kimi-Audio-Inst.2026.06 | 12.7 | 19.2 | |
| AADModel=Kimi-Audio-Inst.2026.06 | 12.3 | 24.8 | |
| AADModel=Voxtral-Small2026.06 | 7.1 | 8.5 | |
| AADModel=Qwen3-Omni2026.06 | 7 | 14.8 | |
| ACDModel=Qwen2-Audio-Inst.2026.06 | 5.6 | 10.4 | |
| ACDModel=Kimi-Audio-Inst.2026.06 | 1.7 | 3.2 | |
| ACDModel=Qwen3-Omni2026.06 | 0.7 | 0.9 | |
| ACDModel=Voxtral-Small2026.06 | 0.5 | 0.5 |